| """Generate original synthetic lab text. Not a foundation-model corpus.""" |
| from pathlib import Path |
| import json |
|
|
|
|
| def main(): |
| rows = [] |
| facts = [ |
| ("code", "A reliable program checks its inputs, handles errors, and returns a well-defined result."), |
| ("code", "Python functions use indentation for blocks. A return statement supplies the result to the caller."), |
| ("systems", "A database transaction groups changes. A rollback restores the state before an unsuccessful transaction."), |
| ("systems", "A queue stores pending work. A bounded queue limits memory use when producers outpace consumers."), |
| ("reasoning", "A hypothesis is a proposed explanation. Evidence can support it or reveal a contradiction."), |
| ("math", "An even integer is divisible by two. The sum of two even integers is even."), |
| ("tools", "A tool receipt records the observed result. A successful request is different from a verified task outcome."), |
| ("privacy", "Personal memory should record its source and confidence. Records must support correction and deletion."), |
| ("hindi", "यह एक छोटा प्रशिक्षण उदाहरण है। सही उत्तर देने से पहले उपलब्ध प्रमाण की जाँच करें।"), |
| ("hinglish", "Code ko run karke output check karo. Agar test fail ho to error samjho aur phir repair karo."), |
| ] |
| for i, (domain, text) in enumerate(facts): |
| rows.append({"id": f"original-{i}", "text": text, "domain": domain, "source": "nexora-original-synthetic-v1", "license": "owner-authored", "split": "train"}) |
| for i in range(80): |
| rows.append({"id": f"function-{i}", "text": f"def offset_{i}(value):\n \"\"\"Add the integer {i} to the input value.\"\"\"\n return value + {i}\n\nThe function offset_{i} maps 3 to {i+3}.", "domain": "code", "source": "nexora-original-synthetic-v1", "license": "owner-authored", "split": "train"}) |
| for i, text in enumerate([ |
| "Validation examples are separate from training examples. Their loss estimates next-token prediction on this small artificial domain only.", |
| "A software test compares observed behavior with an expected result. Failed assertions identify differences that require investigation.", |
| "def multiply(left, right):\n return left * right\n\nThe product of three and seven is twenty-one. Unicode round trips must preserve original input.", |
| ]): |
| rows.append({"id": f"validation-{i}", "text": text, "domain": "validation", "source": "nexora-original-synthetic-v1", "license": "owner-authored", "split": "validation"}) |
| path = Path("examples/corpus.jsonl") |
| path.parent.mkdir(exist_ok=True) |
| path.write_text("".join(json.dumps(r, ensure_ascii=False)+"\n" for r in rows), encoding="utf-8") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
